Munich Personal RePEc Archive
Does land abundance explain African
institutions?
Fenske, James
Yale University
June 2010
Online at https://mpra.ub.uni-muenchen.de/23222/
MPRA Paper No. 23222, posted 11 Jun 2010 05:19 UTC
DOES LAND ABUNDANCE EXPLAIN AFRICAN INSTITUTIONS?
JAMES FENSKE†
Abstract. I show how abundant land and scarce labor shaped African institutions before
colonial rule. I present a model in which exogenous land quality and endogenously evolving
population determine the existence of land rights, slavery, and polygyny. I use cross-sectional
data on pre-colonial African societies to demonstrate that, as in the model, the existence
of land rights, slavery, and polygyny occurred where land was most suitable for agriculture,
and where population density was greatest. These results are robust to alternative measures
of institutions and historical population, and better fit the data than alternative theories of
slavery.
1. Introduction
The “land abundance” view of African history is an influential explanation of the economic
institutions that existed on the continent before colonial rule (Austin, 2008a; Hopkins, 1973;
Iliffe, 1995). This theory holds that, since uncleared land was freely available, land had no
price, rights to land were ill-defined, cultivators would not become free workers, coerced and
household labor substituted for wage employment, capital markets were constrained because
land had no value as collateral, and states that could not tax land remained small and weak.
In this paper, I use a formal model and cross-sectional data on African societies to explain
African institutions. I show how land rights, slavery, polygyny, and state strength in Africa
prior to colonial rule were shaped by the continent’s sparse population.
Institutional failures are a major cause of African poverty. These include corruption, bu-
reaucracy, a lack of democracy, and poor public services (Collier and Gunning, 1999a,b). The
continent’s rare successes, similarly, are understood largely as stories of institutions (Ace-
moglu et al., 2003). This is not unique to Africa, as institutions a principal channel through
†Department of Economics, Yale University, Box 208264, New Haven, CT 06520-8264, Phone:(203) 809-4386, Fax: (203) 432-6323E-mail address: [email protected]: June 10, 2010.I would like to thank my advisors Timothy Guinnane, Benjamin Polak, and Christopher Udry for theirguidance. I would also like to thank Tayo Adesina, Achyuta Adhvaryu, Gareth Austin, Reena Badiani,Benjamin Chabot, Rahul Deb, Shatakshee Dhongde, Nils-Petter Lagerlof, Naomi Lamoreaux, GiuseppeMoscarini, Sheilagh Ogilvie, Mark Rosenzweig, Mir Salim, Richard Smith, Ed Vytlacil, Warren Whatley,Ademola Yakubu, and the participants of the Harvard Economic History Tea, the African Studies AssociationAnnual Meeting, the Queen’s Economic History Conference, the Economic History Association AnnualMeeting, the Canadian Network for Economic History, and the NEUDC for their comments and advice.Thank you as well to Nathan Nunn for sharing Murdock’s map with me.
1
2 JAMES FENSKE
which history affects the present (Greif, 2006; North, 1991; Nunn, 2009). Institutions, in
particular those that protect private property, were instrumental in the rise of the “West”
(Acemoglu et al., 2009, 2005; North and Weingast, 1989) and explain many differences in
outcomes across former colonies (Acemoglu et al., 2001; Banerjee and Iyer, 2005; Dell, 2009).
History has shaped African development. Colonial institutions and investments affect
outcomes today (Bertocchi and Canova, 2002; Price, 2003). Existing arrangements, African
resistance and limited resources, however, constrained colonial powers (Austin, 2008b; Bubb,
2009). As a result, pre-colonial institutions and the forces that shaped them, including states,
polygyny, and slavery, also affect current performance in Africa (Gennaioli and Rainer, 2007;
Nunn, 2008; Tertilt, 2005). Explaining pre-colonial institutions in Africa is, then, important
in understanding its current poverty. Geography is one candidate. Geographic features,
such as continental orientation, ruggedness, settler mortality, suitability for specific crops,
and other biogeographic endowments predict contemporary institutional differences across
countries (Easterly and Levine, 2003; Engerman and Sokoloff, 1997; Nunn and Puga, 2007).
The “land abundance” view of African history argues that the continent’s geography has
given it an abundance of land relative to labor, which explains the general features of its
development. I test this thesis. I use data on a cross-section of African societies from Mur-
dock’s (1967) Ethnographic Atlas to support a model of land rights and slavery in which
the land-labor ratio determines the institutions that exist. I find that the model correctly
predicts that land rights and slavery were found in those societies that occupied the best
land, and that greater population densities were correlated with rights over land. Slavery
was present in the most densely settled parts of Africa, reflecting the high opportunity cost
of coercion at low levels of population and the inability of population to grow to the point
where free labor replaced slavery. Polygyny existed in the most agriculturally suitable and
most thickly settled parts of Africa; dense population is needed for inequality to emerge. This
is consistent with an extension I make to the model, but is a revision to the “land abun-
dance” thesis. While states were more developed in the most populated regions, agricultural
suitability was not one of their systematic determinants.
I provide further evidence in favor of the “land abundance” view by using it to explain
institutions and institutional change among the Egba of southwestern Nigeria between 1830
and 1914. This is an abbreviated form of an analytical narrative I have made in greater detail
in The Author (2010). While the Egba fit many standard predictions for a land-abundant
society, there are two key exceptions. First, they sold land amongst themselves as early
as 1870. Second, land disputes existed. These are explained by initially high population
densities created by their settlement as refugees at Abeokuta, and by the specific features of
certain parcels of land that gave them uncommon value.
LAND ABUNDANCE 3
In Section 2, I outline the literature in African history on how land abundance has shaped
economic institutions. In Section 3, I present the model, extend it to include polygyny, and
identify its testable implications. In Section 4, I describe the data used and lay out the
econometric specifications. In Section 5, I report the results of these tests. In Section 6, I
show that these results are robust to different measures of the institutional outcomes, alter-
native proxies for historical population density, several other interpretations of the results,
and rival theories of slavery. In Section 7 I conclude.
2. The land abundance view of African history
Herbst (2000, p. 16) estimates the population density of Sub-Saharan Africa in 1900 at
4.4 persons per Sq. Km, contrasted with 38.2 for South Asia, 45.6 for China, and 62.9
for Europe.1 Explanations of low African population densities stress geographic factors,
the disease environment, and historical factors such as the slave trades (Mahadi and Inikori,
1987, p. 63-64). This sparse settlement, Hopkins (1973, p. 23-27) argues, shaped institutions,
because Africans “measured wealth and power in men rather than in acres.”2 Iliffe (1995,
p. 1-2) summarizes this “land abundance” view:
Sparse populations with ample land expressed social differentiation through
control over people, possession of precious metals, and ownership of livestock
... Scattered settlement and huge distances hindered transport, limited the
surplus the powerful could extract, prevented the emergence of literate elites
and formal institutions, left the cultivator much freedom, and obstructed state
formation.
Here, I review the literature on how sparse population shaped land, labor, and states in
pre-colonial Africa.3
Before the Atlantic slave trade, Africa was characterized by settled clearings surrounded
by vast wastelands in the Equatorial region, circles of increasingly wild vegetation in the
West African forest, and clusters with oscillating frontiers in the West African Savanna
Iliffe (1995, p. 36, 64-67). Austin (2009b, p. 33) argues that, as a consequence, land was
“easily and cheaply accessible in institutional terms”; pre-colonial authorities were eager to
attract “more people with whom to subdue nature and, if necessary, their neighbors,” so
that strangers could generally acquire land indefinitely for token payments. These payments
were made solely to acknowledge the sovereignty of the local authorities. Citizens were given
land virtually freely. Austin (2008a, p. 591-594) notes that ‘islands’ of intensive agriculture
1His estimate for North Africa is 9.4 persons per Sq. Km.2Austin (2008a, p. 589) argues that Hopkins was the first to make this analysis systematic; earlier writers onAfrica did account for the existence of slavery, for example, by noting Africa’s land abundance – see Dowd(1917).3Capital markets are another major theme of this literature, but the data that I have do not permit anystatistical tests. I examine Egba credit institutions in The Author (2010).
4 JAMES FENSKE
have existed where insecurity has created artificial land scarcity and in specific locations
of exceptional value. These had minerals, trees, market access, or suitability for particular
crops.
Against these views, Spear (1997, p. 154-157) argues that population density cannot ex-
plain individual cases. While on Mount Meru both the Arusha and the Meru intensified
their agriculture as population rose, the less densely settled Meru did so more readily. Berry
(1988), similarly, has noted that inheritance rules, tenancy contracts, and labor arrange-
ments often prevent tree crops from leading to individualized land tenure in West Africa.
Thornton (1992, p. 75-76) suggests that ownership of land results from legal claims, not
population pressure. In Section 5, I show that the institutional effects of population and
agricultural productivity are systematic, even if they are not deterministic.
For Austin (2008a, p. 606-610), scarcity of labor explains African use of extensive agri-
culture, dry season crafts and industries, and forced labor. With some notable exceptions
(Rodney, 1966), slavery was prevalent in much of Africa even prior to the Atlantic slave
trade (Fage, 1969). Watson (1980, p. 10) suggests that the ability of slaves and their descen-
dants to assimilate into their owners’ lineages was a “logical extension of the institutionalized
need for more people.” Land abundance has been used to explain differences across soci-
eties. Northrup (1979) contrasts the densely-settled Igbo of the palm belt with the relatively
sparsely populated northeastern Igbo during the palm oil trade. Slavery did not expand in
the palm belt, while the northeastern Igbo used slaves to colonize new land.
Family structures in Africa have also been linked to sparse population. Tambiah and
Goody (1973, p. 23) explain bride-price by noting that, since men are not distinguished by
land holdings, the price of a husband is low. Iliffe (1995, p. 96) argues that intense compe-
tition for women within and across generations led to the payment of bridewealth. Because
wives’ labor and reproductive capacities are so important, more than half of customary court
cases in Africa are disputes over marriage, divorce or bridewealth (Kopytoff, 1987, p. 43). I
argue instead that polygyny can only exist when population is great enough for an elite to
have already differentiated itself from the population.
The use of underpopulation to explain African slavery is controversial. Kopytoff and
Miers (1977, p. 68-69) object that slaves filled social and political functions for which entire
persons were needed, and not simply their labor. Political insecurity prevented people from
taking advantage of surplus land. Lovejoy (1978, p. 349) argues that slavery in the Sokoto
Caliphate was “based on non-market principles,” as slaves and output were redistributed
mostly through the state. Miers and Klein (1998, p. 4-5) and Roberts and Miers (1988,
p. 20) stress factors other than labor scarcity that made colonial rulers hesitant to abolish
slavery, including their dependence on slave-owners, fear that abolition would divert trade,
worries about disrupting peace, unwillingness to undermine male control of women, and their
LAND ABUNDANCE 5
experience with India. Austin (2009a) responds that the rise in slave-holding throughout the
Atlantic slave trade and the nineteenth century cannot be explained by the non-economic
uses of slaves. I show that the presence of slavery across African regions was systematically
related to the economic value of slaves and to population. Kopytoff (1987, p. 46) and Goody
(1980, p. 26-31) suggest that dependents must be “seduced” rather than coerced, so slavery
can only exist in complex societies and states with “well-developed systems of compulsion.” I
show that high opportunity costs of coercion at low population densities can be incorporated
into a model in which slavery is explained by the high cost of free labor.
Prior to colonial rule, the “typical” Atlantic African lived in a state with an area no larger
than 1,500 square kilometers and fewer than 30,000 inhabitants (Thornton, 1992, p. 105).
African states were, Austin (2004a, p. 25) argues, “webs of relationship which grew steadily
weaker with distance from the capital until they merged into the statelessness of peripheral
peoples.” States could not raise revenues from land. Unable to tie subjects to the land and
tax them, states could not make land artificially scarce. Revenues came from other sources,
such as trade tolls. Rulers sought subjects and cattle, rather than territories (Austin, 2004b).
I demonstrate that state strength in Africa has been systematically related to population,
but not to agricultural suitability.
3. Model
In this section, I formalize the literature outlined in Section 2. I use an extension of
the model of “slavery and other property rights” from Lagerlof (2009), adding both slave
raiding from neighboring societies and polygyny. Elite preferences over the three institutional
regimes of egalitarianism, slavery, and free labor are driven by agricultural productivity and
population size. Population lowers wages and average product, making free labor preferable
to slavery or egalitarianism. Productivity makes coercion worthwhile. This adds to the
literature by recognizing the importance of productivity and the opportunity cost of coercion
when population is low. Section 3.1 sets up the model. Section 3.2 describes its dynamics.
Section 3.3 adds polgyny. Section 3.4 derives two testable implications of the model. Greater
suitability for agriculture will positively predict the existence of land rights, slavery and
polygyny. Endogenous population density will be positively correlated with land rights and
polygyny, positively correlated with slavery if population growth is limited by the disease
environment, and non-monotonically correlated with slavery if population growth is not so
constrained.
3.1. Setup. A society in period t has a population Pt of non-elite agents and a population
of elite agents that has zero mass. Non-elite agents work; elite agents do not. Both live
for one period. The elite is randomly selected from the population at the beginning of each
6 JAMES FENSKE
period. Agents choose fertility nt and consumption ct. Children cost q each. If income is It,
each agent’s budget constraint is:
ct = It − qnt.(1)
With no utility from leisure, non-elite agents supply one unit of labor each. Utility is:
Ut = (1− β) ln ct + β lnnt.(2)
This implies that optimal fertility n∗
t for each agent is given by:
n∗
t = (β/q)It.(3)
Output Yt depends on land T , land-augmenting productivity A, and the labor used Lt:
Yt = (TA)αL1−αt ≡ AαL1−α
t ,(4)
where α ∈ (0, 1). A depends on A and T , but is interpreted here as agricultural suitability,
given exogenously by the environment. The elite’s payoff in period t under each of the three
institutions, egalitarianism, slavery, and free labor, is given by πit, where i ∈ E, S, F. The
population’s payoff is mit. At the beginning of each period, the society’s neighbors raid it for
R slaves, and nothing can prevent this. There is no voluntary migration; agents who leave
will be enslaved by their neighbors.
3.1.1. Egalitarianism. Under egalitarianism, there are no land rights or slavery. The elite
and the non-elite agents that remain after the society is raided receive average product:
πEt = mE
t =( A
Pt −R
)α
.(5)
3.1.2. Free labor. Under free labor, the elite encloses a fraction θ of the land, creating rights
over it. To do this, they require the support of an infinitely-lived external elite of mass 1.
They share their income equally with this external elite, though members of the external
elite do not make decisions or have children.4 They hire Lt non-elite agents to work for them
at a competitive wage wt. The elite’s problem is:
πFt = max
Lt∈[0,Pt−R](θA)αL1−α
t − wtLt.(6)
4This is done solely so that the elite’s vanishingly small income under egalitarianism is comparable to itsfinitely positive income under free labor or slavery.
LAND ABUNDANCE 7
Non-elite agents not hired continue to work the remaining land communally, receiving
income mFt =
(
(1−θ)APt−R−Lt
)α
. Equilibrium is achieved in the labor market when the wage
(equal to the marginal product of labor on the elite’s estate) is equal to the average product
on the unenclosed land. This will be true when:
(1− α)(θA)αL−αt =
( (1− θ)A
Pt −R− Lt
)α
,(7)
which implies that the optimal choice of labor for the elite, L∗
t , is given by:
L∗
t =(1− α)
1
α θ
(1− θ) + (1− α)1
α θ(Pt −R) ≡ σ(Pt −R),(8)
which in turn implies that the equilibrium wage (and hence non-elite income) is given by:
wt = mFt = (1− α)(θA)α(σ(Pt −R))−α.(9)
Substituting (8) and (9) into (6), the elite’s payoff is:
πFt = αθασ1−αAα(Pt −R)1−α.(10)
3.1.3. Slavery. Under slavery, the elite again uses the help of the external elite to enclose
a fraction θ of the land, creating rights over it. They raid their neighbors for slaves, at an
elastic cost r,5 which includes the cost of guarding the slaves and feeding them while they
are used in production. It is assumed for simplicity that free workers will not work alongside
slaves. Slaves do not reproduce.
The elite’s problem is:
πSt = max
St
(θA)αS1−αt − rSt.(11)
Solving for the elite’s preferred number of slaves, the elite’s payoff is:
πSt = α
[1− α
r
]1−α
α
θA.(12)
The non-elite population receives the average product on the unenclosed land:
5It is assumed the elite’s holding is small enough relative to its neighbors’ population that it does not facethe possibility of enslaving the entire neighboring population.
8 JAMES FENSKE
mSt =
((1− θ)A
Pt −R
)α
.(13)
3.1.4. Comparing payoffs. The elite chooses the institutional regime that suits them the most
in any particular period. Because they are only lived for one period, they are not forwards-
looking. Their preferences will be determined by the level of agricultural suitability, A, and
by the population that period, Pt. In comparing their payoffs under each regime, is is helpful
to define the following functions of Pt:
Ψ(Pt) =[ r
1− α
]1/ασ
θ(Pt −R).(14)
Ω(Pt) =( 1
αθ
)1
1−α
( r
1− α
)1
α
(Pt −R)−α
1−α .(15)
Φ =1
αθασ1−α.(16)
These partition the (A,Pt) space into three sets:
SE = (A,Pt) ∈ R2+ : (A,Pt) /∈ SS ∪ SF,
SS = (A,Pt) ∈ R2+ : A ≥ maxΨ(Pt),Ω(Pt),
SF = (A,Pt) ∈ R2+ : Pt ≥ Φ +R and A ≤ Ψ(Pt).
(17)
These regions are depicted in Figure 1. These define the elite’s institutional preferences:
Proposition 1. Elite preferences over institutions are determined by A and Pt:
I. Egalitarianism is weakly preferred when:
πEt ≥ maxπS
t , πFt ⇔ (A,Pt) ∈ SE.
II. Slavery is weakly preferred when:
πSt ≥ maxπE
t , πFt ⇔ (A,Pt) ∈ SS.
III. Free labor is weakly preferred when:
πFt ≥ maxπE
t , πSt ⇔ (A,Pt) ∈ SF .
Proof. (5) and (12) imply that πSt ≥ πE
t iff A ≥ Ω(Pt). (10) and (12) imply that πSt ≥ πF
t iff
A ≥ Ψ(Pt). (5) and (10) imply that πFt ≥ πE
t iff Pt ≥ Φ +R.
Slavery is preferred when population is large enough that the opportunity cost of coercion
is low, but small enough that free labor is expensive in comparison. Greater agricultural
LAND ABUNDANCE 9
productivity overcomes the inefficiency of coercion. Population growth pushes down the
average product of land, making egalitarianism unattractive.
Figure 1. Institutional regions and dynamics
𝐴
𝑃𝑡
Ψ(𝑃𝑡)
Φ+ 𝑅
Ω(𝑃𝑡)
𝒮E
𝒮S
𝒮
𝒮F
A
𝐋𝐄(𝑃𝑡)
𝐋𝐅(𝑃𝑡) 𝐋𝐒(𝑃𝑡)
3.2. Dynamics. Population evolves according to:
Pt+1 = nnon-elitet (Pt −R).(18)
Using (3), (5), (9), (13), and (18), population is constant when:
A =
(
qβ
)1
α
(
Pt
(Pt−R)1−α
)1
α ≡ LE(Pt) if (A,Pt) ∈ SE
(
qβ(1−α)
)1
α σθ
(
Pt
(Pt−R)1−α
)1
α ≡ LF (Pt) if (A,Pt) ∈ SF
(
qβ
)1
α 11−θ
(
Pt
(Pt−R)1−α
)1
α ≡ LS(Pt) if (A,Pt) ∈ SS
(19)
10 JAMES FENSKE
If A > LS(Pt) under slavery, Pt is rising. If A < LS(Pt) under slavery, Pt is falling. If
A > LE(Pt) under egalitarianism, Pt is rising. If A < LE(Pt) under egalitarianism, Pt is
falling. If A > LF (Pt) under free labor, Pt is rising. If A < LF (Pt) under free labor, Pt is
falling. These are simple Malthusian “zero population growth” lines; if A is large relative to
Pt, income is high, and population is growing. If A is low, income is low, and population is
falling. The lines themselves depend on the institutional region, because income is shared
differently under each regime.
Again, it is helpful to define the levels of A at which these zero population growth lines
intersect the borders of the institutional regions:
AΦΨ,Ω =
( r
1− α
)1
α
( σ
θΦ
)
,(20)
AFΦ =
( q
β(1− α)
)1
α σ
θ
(Φ +R
Φ1−α
)1
α
(21)
ASΦ =
( q
β
)1
α 1
1− θ
(Φ +R
Φ1−α
)1
α
(22)
AΦΨ,Ω is the level of A at which Ψ(Pt) and Ω(Pt) intersect Pt = Φ + R. AF
Φ is the level of
A at which LF (Pt) intersects Pt = Φ + R. ASΦ is the level of A at which LS(Pt) intersects
Pt = Φ+R. The dynamics in (19) determine what steady states will exist:
Proposition 2. Steady states.
I. So long as A is below a cutoff AE(α, β, θ, q, r, R), there is a steady state under egalitar-
ianism.
II. If AFΦ ≤ AΦ
Ψ,Ω and A is above a cutoff AF (α, β, θ, q, r, R), then a steady state under free
labor may exist.
III. If ASΦ ≥ AΦ
Ψ,Ω and A is above a cutoff AS(α, β, θ, q, r, R), then there is a steady state
under slavery.
Proof. So long as A is low enough, it will obviously intersect LE(Pt) in SE. AFΦ ≤ AΦ
Ψ,Ω
ensures LF (Pt) is flat enough to intersect SF . If AF is chosen as the level of A at which
LF (Pt) intersects Pt = Φ + R, above this the intersection of A and LF (Pt) may occur in
SF . Finally, ASΦ ≥ AΦ
Ψ,Ω ensures LS(Pt) is steep enough to intersect SS. If AS is chosen as
the level of A at which LS(Pt) intersects Ω(Pt), for any A ≥ AS, the intersection of A and
LS(Pt) will occur in SS.
An example with a steady state under free labor is depicted in Figure 1.
3.3. Polygyny. Assume now that “wives” are an input into the production of children.
Following Tertilt (2005), the cost of producing nt children using Wt wives is now qn2t/Wt.
LAND ABUNDANCE 11
If the purchase price of a wife is bt, the total cost of nt children borne by Wt wives will
be btWt + qn2t/Wt. This captures the idea that, while wives must be purchased, the cost of
additional children is convex for any particular wive, due (for example) to maternal mortality
or time constraints. Each member of the elite and of the non-abducted non-elite population
has h sisters who he sells at the market price of bt. With a balanced sex ratio, h would equal
1. Payment of bride price to the brother simplifies the model by removing receipt of bride
price as a motivation for fertility.
The cost-minimizing choice of Wt for an agent taking bt as given will be√
(q/bt)nt. The
cost-minimizing choice of wives implies that there is a linear marginal cost of children equal
to 2√btq. If an agent has income It, his optimal number of children will be βIt/(2
√btq),
which implies that his demand for wives is βIt/(2bt). Under institution i, elite income is
πit + bth, while non-elite income is mi
t + bth. Since the elite has zero mass, total demand in
the market for wives is:
β
2bt
[
mit + hbt
]
(Pt −R).(23)
The total supply of wives is given by h(Pt−R). In equilibrium, bt is set by the intersection
of total demand with total supply, where:
bit =β
(2− β)hmi
t.(24)
Polygyny exists when the elite has more wives than members of the population, i.e when
β(πit + bth)/(2bt) > β(mi
t + bth)/(2bt) or πit > mi
t. The addition of polygyny will not
qualitatively change the elite’s preferences over institutions.6 Under egalitarianism, the
elite’s income will be the same as that of the non-elite, and so polygyny will not exist.
Under free labor, the condition that Pt ≥ Φ + R ensures that the elite’s income will be
greater than that of the non-elite, and so polygyny will exist. Under slavery, the condition
that A > Ω(Pt) similarly ensures that the elite has more wives. Inequality is a precondition
for polygyny. This is a revision of the “land abundance” interpretation of African history.
6Under institutional setting i, if the elite receives income πit and the equilibrium bride price is bit, the elite’s
maximized utility will be equal to:
V it = (1− β) ln
(
πit − 2
√
bitqn∗t
)
+ β ln(n∗t ) = ln(πi
t)−β
2ln(bit) +K,
where K is a constant. Thus, institution i will be preferred to institution j if V it ≥ V j
t , or:
ln(πit)−
β
2ln(bit) ≥ ln(πj
t )−β
2ln(bjt ) ⇒
πit
πjt
≥( bit
bjt
)
β
2
.
From (24), bit/bjt = mi
t/mjt . The ratios mi
t/mjt are constants independent of A and Pt. While the definitions
of Φ, Ω, and Ψ must be adjusted to include these constants, their general shapes will not change.
12 JAMES FENSKE
3.4. Tests. Two implications of the model are tested in Section 5:
I. Increasing exogenous agricultural suitability (A) predicts the existence of slavery, polyg-
yny and rights over land.
II. Polygyny and land rights exist when endogenous population density (Pt) is high. If there
are no constraints on population growth, slavery exists at intermediate Pt. If population
growth is limited, slavery will be positively correlated with population density.
Rights over land and polygyny exist under both slavery and free labor. A ≥ AF is
necessary for a steady state to exist under free labor. Since AF is a nonlinear function of
model parameters that are not observed, a matrix of geographic controls X is used to proxy
for AF by assuming:
AF ≈ 1
δ0(−X ′λ0 − ǫ0),(25)
where δ0 and λ0 are regression coefficients and ǫ0 is an error term. The probability that a
steady state exists under free labor (i.e. with land rights and polygyny) is:
Pr(Steady state in SF ) = Pr(ǫ0 ≥ −δ0A−X ′λ0).(26)
If ǫ0 ∼ N(0, 1), this can be estimated as a probit. Similarly, A ≥ AS is necessary for a
steady state to exist under slavery. The probability of a steady state with slavery is:
Pr(Steady state in SS) = Pr(ǫ1 ≥ −δ1A−X ′λ1).(27)
Again, if ǫ1 ∼ N(0, 1), this can be estimated as a probit.
Land rights and polygyny exist under free labor and slavery, i.e. when Pt ≥ minΦ +
R,Ω−1(A) = minΦ + R,Ω−1(A,P0, t). Again using X as a proxy, the probability that
land rights or polygyny exist for an observed A and Pt is:
Pr(Land Rights, Polygyny) = Pr(ǫ2 ≥ −δ2iPt − δ2iiA−X ′λ2).(28)
If ǫ2 ∼ N(0, 1), this can be estimated using a probit. According to the “land abundance”
view of African history, population could not grow to the point where free labor replaced
slavery, and so this will also be the condition for the existence of slavery. Without this
restriction on population, if A is large enough, slavery will exist when Ψ−1(A) ≥ Pt ≥Ω−1(A). Using X, this is equivalent to stating that slavery exists if:
δ3iPt + δ3iiA+X ′λ3 + ǫ3 ≥ 0 and δ4iPt + δ4iiA+X ′λ4 + ǫ4 ≥ 0.(29)
LAND ABUNDANCE 13
If (ǫ3,−ǫ4) ∼ N(0,Λ), this is the Poirier (1980) partially unobserved bivariate probit
model. However, because this could not be implemented on the actual data, the tests used
look for an inverted-U relationship between population density and slavery.7
4. Data and Specifications
In this section, I outline how I test the two predictions of the model described above. I
use a cross section of data on 531 African societies, observed on the eve of colonial rule.
In Section 4.1 I detail the specific econometric specifications that I use. In Section 4.2, I
describe the sources of data on institutions, the proxies for the variables A and Pt in the
model, and the additional controls that I include. In Section 4.3, I describe the historical
sources that are used to provide supporting detail on Egba institutions and their evolution.
4.1. Specifications. The first prediction of the model is that raising A will make it possible
for steady states to exist with land rights, polygyny, or slavery. I test this by estimating:
yi = δr + βAAi +X ′
iγ + ǫi,(30)
where yi is an outcome of interest for ethnic group i, δr is a vector of dummies for the
fifteen regions in the data (described below), Ai is a proxy for agricultural suitability, Xi is
a matrix of geographical controls, and ǫi ∼ N(0, 1) is random error. (30) is estimated as a
probit, and observations are weighted by estimated population. This is done to avoid giving
smaller groups undue influence in the results. Standard errors are clustered by region. I
expect that βA > 0 for land rights, slavery, and polygyny.
The second implication of the model is that land rights and polygyny exist at higher levels
of Pt, while slavery exists at intermediate levels of Pt. I test these by estimating:
yi = δr + βP ln(Pi) + βAAi +X ′
iγ + ǫi,(31)
and
yi = δr + βP1Pi + βP2P2i + βAAi +X ′
iγ + ǫi,(32)
where yi, δr, Ai, Xi, and ǫi are defined as in (30). Pi is the proxy used for Pt. These are
also estimated as probit models, with observations weighted by estimated population and
standard errors clustered by region. I expect that βP > 0 for land rights, polygyny, and
7There are no elements of X that can be excluded a priori from either of the two equations in the partiallyunobserved bivariate probit model. Without an exclusion restriction of this type, the model may not beidentified on actual data, as is the case with the data used below.
14 JAMES FENSKE
slavery in the restricted model, and that βP1 > 0 and βP2 < 0 for slavery in the unrestricted
model.
Finally, I test for neighbor effects by estimating a spatial autoregressive (SAR) model:
yi = α + ρWiyi−1 + βAAi +X ′
iγ + ǫi.(33)
Here, α is a constant and Wi is an N × N spatial weight matrix, in which each entry
Wij is an indicator for whether observation i borders observation j, normalized so that its
rows sum to 1 or 0. yi−1 is a vector of outcomes for the other observations. ρ captures
whether the institutional outcome of one group will affect its neighbor’s institutions. In the
model, this could operate through many parameters. For example, if a society’s neighbor
shifts into the slavery region, R, the number of slaves raided, will likely increase. ρ is not
separately identified from localized unobservables. However, not all estimates of ρ are found
to be positive, which suggests that the spillovers found are not due solely institutional shocks
common to neighboring societies. Because of the spatial lag, standard probit estimates will
be inconsistent. The model is estimated using the Markov Chain Monte Carlo SAR probit
estimator described by LeSage and Pace (2009, p. 283-289).
4.2. Data. Two types of data are used to test the ability of the model to explain institutional
differemces across societies within Africa. The first covers institutions, and is taken from
Murdock’s (1967) Ethnographic Atlas. Published in 29 installments of the journal Ethnology
between 1962 and 1980, the Atlas is a database of 1267 societies from around the world.8 It
contains categorical variables describing several institutional and cultural features of these
societies at the time of first contact with Europeans. 531 African societies are used for the
analysis.9
Four variables from the Ethnographic Atlas are used to construct binary dependent vari-
ables, and summary statistics for these are given in Table 1.10 Indicators are used for whether
individual land rights, slavery, or polygyny exist. The measure used of state power is whether
8A revised version of the Atlas has been made available for download in SPSS format by J. Patrick Grayat http://eclectic.ss.uci.edu/˜drwhite/worldcul/EthnographicAtlasWCRevisedByWorldCultures.sav. Thisis the version used for the present study.9The Guanche, an extinct people of the Canary Islands, are dropped because they are observed more than300 years earlier than any of the other groups in the African sample barring Ancient Egypt, which is similarlydropped. Dates of observation are missing for the Bomvana and Betsileo. The Bomvana are recoded to 1850,to match the date of observation for the other Xhosa, while the Betsileo are recoded to 1900, the modal datefor the other Malagasy societies in the data.10These are: V74: Inheritance Rule for Real Property (Land); V70: Type of Slavery; V9: Marial Com-position: Monogamy and Polygamy; and V33: Jurisdictional Hierarchy Beyond Local Community. Thedefinitions of the binary variables are: 1) Land rights exist if V74 6=1, 2) slavery exists if V70 > 1, 3)polygyny exists if V96=1, and 4) state centralization exists if V33>2.
LAND ABUNDANCE 15
Table 1. Summary statistics
MeanStd.Dev. Min Max N
Institutional Outcomes
Any Individual Land Rights 0.93 0.25 0 1 404Any Slavery 0.85 0.36 0 1 454Polygyny 0.95 0.21 0 1 517State Stratification 0.34 0.47 0 1 475
Alternative Institutional Outcomes
Patrilineal Land Rights 0.77 0.42 0 1 404Consideration for Bride 0.93 0.26 0 1 529Class Stratification 0.53 0.50 0 1 426
Geographic Controls
Agricultural Suitability 0.53 0.18 0 1 531Population Density (1960) 21.8 29.3 0 315 531Elevation 703 506 -14.9 2306 531Precipitation 1.12 0.57 0.013 2.98 531Temperature 8.82 1.19 5.31 10.8 531Malaria Suit. 0.77 0.33 0 1 531Tsetse Suit. 0.54 0.42 0 1 531Dist. to Coast. 5.50 3.84 0.023 14.9 531Dist. to Lake Victoria 2.37 1.51 0.13 5.8 531Ruggedness 0.22 0.08 0.031 0.77 531Abs. Latitude 9.89 7.58 0.017 36.6 531Population Weight 410 1267 0.34 25611 531
Controls from Ethnographic Atlas
Major Crop: Missing 0.07 0.26 0 1 531Major Crop: None 0.02 0.15 0 1 531Major Crop: Tree Fruits 0.09 0.28 0 1 531Major Crop: Roots andTubers
0.16 0.37 0 1 531
Date Observed 1919 21.7 1830 1960 531
Notes: The omitted crop type is cereal grains, the mode.
there is more than one level of jurisdiction above the local. The model gives clear predic-
tions for land rights and slavery, and there is adequate variation in these measures for stable
econometric tests. For polygyny, however, the lack of variation in the data makes the results
of econometric tests sensitive to the specification chosen. Results for polygyny, then, will be
suggestive but not dispositive. The model says nothing about states, but these are included
as an outcome because they are central to the “land abundance” view.
16 JAMES FENSKE
The second type of data used includes features of the natural environment. These are
joined to the data from the Ethnographic Atlas using the “Tribal Map of Africa” from
Murdock (1959). Sources and definitions of these variables are given in Table 10 in Appendix
A. The first step of this join requires matching ethnic groups in the Atlas to those in the
Map. This was done first by name, then by location. The majority (426) were matched
exactly by name, and most of the rest were matched by using an alternative spelling (40)
or alternative name (15). For some, the division of ethnic groups in the atlas did not match
that in the map, and so these were matched either to a larger group of which they are are
a part (a “supergroup” – 20), a smaller group (a “subgroup” – 4), or another group that is
part of the same “supergroup” (an “alternative supergroup” – 5). Finally, 21 groups could
not be identified with those in the map, and so were matched to the group that is at the
latitude and longitude co-ordinates specified in the Atlas. Table 11 in Appendix A reports
how groups that did not match exactly by name were joined. This table also includes an
ISO 693-3 code that indicates a corresponding entry in Gordon and Grimes (2005). Where
this entry does not contain enough information on its own to justify the match, additional
notes have been added to Table 11.
The second step in this merge involves joining geographic raster data to Murdock’s (1959)
map by taking the average of the points within an ethnic group’s territory. Summary sta-
tistics for these variables are presented in Table 1. Two of these controls are of particular
importance – agricultural suitability and population density.
4.2.1. Agricultural suitability. The variable used to capture agricultural suitability is based
on Fischer et al.’s (2002) measure of combined climate, soil and terrain slope constrains. This
is re-scaled to lie between 0 and 1, with larger values indicating an absence of environmental
constrains on rainfed agriculture. This is treated as a proxy for the variable A in the model.
The constraints measure was constructed as part of the Food and Agriculture Organiza-
tion’s Global Agro-Ecological Zones (FAO-GAEZ) methodology. This methodology combines
multiple sources of data on climate, soils, and landform to quantify the expected productivity
of all feasible land use and management options on a global scale. The constraints measure
is not particular to any particular crop or technology, and is a non-additive combination of
three components:
(1) Climate constraints: The coldness constraint is “moderate” if there are fewer than
180 days with an average temperature below 5C, and “severe” if there are fewer
than 120. Aridity constraints are moderate if there are less than 120 days with an
average temperature below 5C during which moisture conditions are adequate to
permit crop growth and severe if there are less than 60.
(2) Soil constraints: Five characteristics of soils are considered – depth, fertility, drainage,
texture and chemical constraints. “Medium” and “shallow” depth are moderate and
LAND ABUNDANCE 17
severe constraints, respectively. “Medium” and “low” fertility are treated similarly
as moderate and severe constraints. “Poor” drainage is a severe constraint. Sandy
and stony soils are severe constraints, and cracking clay is a moderate constraint.
Salinity, sodicity, and gypsum are severe chemical constraints.
(3) Terrain slope constraints: Terrain slopes greater than 8% are “moderate” constraints,
and slopes greater than 30% are “severe.”
Climate constraints and soil texture are clearly exogenous. Given the low level of pre-
colonial agricultural technology, it is unlikely that terrain slope, drainage, and chemical
constraints are consequences of institutions. It is possible that societies that developed
slavery or rights over land were able to avoid degrading the soil depth and fertility. Since
these are only two components of a larger measure, the bias should be small. In addition, the
direct measures of soil depth and fertility constraints can be added as additional controls.
All results for agricultural suitability are robust to the inclusion of soil depth. Soil fertility
constraints do, however, eliminate the direct effect of agricultural suitability on land rights.
A caveat must be added, then, that the relationship between agricultural suitability and
land rights may be overstated because land rights create incentives to preserve soil fertility.
4.2.2. Population density. The second important control is population density in 1960, pub-
lished by the United Nations Environment Programme. This is treated as a proxy for Pt
in the model. These data are the work of Nelson and Deichmann (2004), who construct
population measures from official censuses, yearbooks, gazetteers, area handbooks and other
country studies. These data are then interpolated to standardized years using intercensal
growth rates. The use of population density from 1960 as a proxy for pre-colonial densities
is reasonable insofar as the relative distribution of population within Africa has been stable
over time across regions as large as those used as observations. Population density in 1960
and 2000 have a correlation coefficient of 0.92 and their logs have a correlation coefficient of
0.97, which suggests that this is a fair assumption. In the analysis, I also account for the
date of observation, which will capture growth effects. In addition, population growth rates
between 1960 and 2000 are used to back-project population densities in Section 6, and the
results do not change substantially. The measures of agricultural suitability and population
density are plotted together in Figure 2.
4.2.3. Other controls. The other controls listed in Table 1 are included as proxies for the
unobserved cutoffs described in Section 3.4. These are nonlinear functions of α, β, q, r, and
θ. Elevation is related to the disease environment, and hence the cost of children (q). It also
affects the range of available crops and technologies, and hence α. McCann (1999, p. 38-39),
for example, notes that the Ethiopian highlands were a unique source of crops such as teff
and supported both animal husbandry and use of the plough. Precipitation determines what
18 JAMES FENSKE
Figure 2. Agricultural Suitability and Population Density, 1960
Agricultural suitability is on the left, population density on the right. Darker colors indicate higher values;the range of agricultural suitability is from 0 to 1, while the range for population density is from 0 to 315.25.
crops can be grown, shaping α. African growing seasons and diseases are constrained by the
seasonal availability of moisture (McCann, 1999, p. 15-18). Areas with low rainfall are also
those most susceptible to drought (Bloom and Sachs, 1998, p. 222); β and r accounting for
storage needs will be greater.
Temperature affects the physical cost of effort, and hence r and β. In hostile environments,
it is more difficult for slaves to flee; r is lower.11 Temperature affects q through nutrition and
disease (Bloom and Sachs, 1998, p. 228). Distances from the coast and from Lake Victoria
proxy for water-borne diseases that affect q. These distances also capture the presence of
trade, which affects both α and β through what goods are bought and sold, and the cost of
slavery (r) through what uses exist for slaves and whether they can be punished by sale for
export.
The suitability of the environment for malaria affects q through child mortality and r via
slave mortality. It may also alter the physical cost of effort in adults (Gallup and Sachs, 2001,
p. 94-95). Suitability for tsetse makes the survival of draught animals and cattle difficult,
shaping α. Kjekshus (1977, p. 51) writes that the “overwhelming feature in the study of
cattle-keeping in East Africa is the presence of the tsetse fly.” Trypanosomiases also affect
q via human mortality, as well as the ability to use cavalry (and thus r). Webb (1995) cites
this as a decisive factor in the history of the Western Sahel. Ruggedness, following Nunn and
11Isaacman et al. (1980, p. 598) makes a similar point in discussing the difficulties faced by refugees who fledcolonial rule in northern Mozambique.
LAND ABUNDANCE 19
Puga (2007), is related to the cost of capturing slaves, and hence r. Crop dummies are taken
as exogenous determinants of the available technologies (α). Absolute latitudes north and
south of the equator and the date at which the society was first observed are also included
as controls for unobservable heterogeneity.
4.3. Historical sources. In order to supplement the econometric tests, I have also collected
information on the Egba of southwestern Nigeria from their arrival at the town of Abeokuta
in 1830 to their loss of formal independence in 1914. During the period of study, the Egba
cultivated maize, cotton, yams, cassava and beans, supplementing these with other crops,
and exported oil and kernels gathered from wild palm trees to European markets. Late
in the century, cocoa and kola were introduced. These were both tree crops that were
planted intentionally. I take data from secondary sources, oral histories, missionary records,
travelers’ descriptions, official correspondence and private letters. My principal sources are
a collection of 541 Native Court cases involving farmland that took place between 1902 and
1919; these are housed in the National Archives, Abeokuta, and in the Hezekiah Oluwasanmi
Library at Obafemi Awolowo University, Ile-Ife. The testimonies of the participants of these
disputes reveal many details about how land has been used, what transactions have occurred,
what rights exist over the land, and and what the causes are of land conflicts. For a fuller
description and transcribed example, see The Author (2010).
5. Results
In this section, I implement both tests of the model described in Section 3.4. The model
correctly predicts the existence of land rights, slavery and polygyny where A is highest, and
land rights and polygyny are positively correlated with population density. Slavery, however,
is positively related to population density – it does not exist only at intermediate levels. This
may be understood within the model as due to Africa’s disease environment or to its overall
sparse population, as discussed below. I describe the results by outcome, before turning to
the spatial autoregressive model.
5.1. Land Rights. Table 2 presents the results for land rights. As predicted by the model,
an increase in A predicts the existence of rights over land. When only the proxy Ai and
a dummy for the North and Saharan regions (taken together) are included, the estimated
coefficient on Ai is positive and significant. As more controls are included in Column (2),
the estimated coefficient on agricultural suitability becomes larger. 36 observations are lost,
because all societies for which trees are their principal crop have land rights. This is itself
interesting. It supports the argument that agricultural suitability, in this case conditions
favorable for tree crops, predicts the existence of land rights. Further, it fits with the “land
abundance” view that, as embodied labor, tree crops are valuable assets over which rights
were more valuable than rights over land in general. The increase in the magnitude of the
20 JAMES FENSKE
Table 2. Tests of the model: Land rights
(1) (2) (3) (4) (5)
Any individual land rights
Agric. Suitability 2.12*** 3.06** 3.23*** 2.94*** 2.50**(0.740) (1.539) (0.936) (1.019) (0.976)
Ln(Pop. Density) 0.55***(0.143)
Pop. Density (1960) -2.56(4.322)
Pop. Density Sqd. 12.92(8.531)
Observations 404 368 321 321 321Other Controls No Yes Yes Yes YesRegion F.E. North/Sahara No Yes Yes Yes
Notes: ***Significant at 1%, **Significant at 5%, *Significant at 10%. Robust standard errors in parenthe-ses. All regressions are probit, with coefficients reported. Observations are weighted by estimated popula-tion in 1960 and standard errors are clustered by region. Region FE: African Hunters, South African Bantu,Central Bantu, Northeast Bantu, Equatorial Bantu, Guinea Coast, Western Sudan, Nigerian Plateau, East-ern Sudan, Upper Nile, Ethiopia/Horn, Moslem Sudan, Sahara, North Africa, and Indian Ocean. OtherControls: Malaria suitability, tsetse suitability, ruggedness, dummies for major crop types (missing, none,tree fruits, roots/tubers included, cereal grains excluded), date of observation, absolute latitude, abso-lute latitude X latitude > 0, and quadratics in elevation, annual precipitation, accumulated temperature,distance to lake Victoria, and distance to the nearest coast.
coefficient on agricultural suitability suggests that its significance is not due to correlation
of agricultural suitability with other unobservable variables that make the existence of land
rights more likely. If these unobservable features have similar correlations with agricultural
suitability as the observable variables, including them would strengthen the estimated effect
of agricultural suitability on land rights (Altonji et al., 2005). Similarly, the results are robust
to the inclusion regional dummies in Column (3). The marginal effects, while present, are
small; in Columns (1) and (3), a one standard deviation increase in agricultural suitability
raises the probability of land rights by roughly 2%.12
The results concerning Pt in Column (4) also confirm the predictions for land rights. This
is not a causal effect; as an endogenous variable, it is correlated with land rights as the
model predicts. The marginal effect here suggests that a one standard deviation increase
in log population density is associated with a 2% increase in the likelihood of land rights
existing. Column (5) has no meaning in this table; it is reported for completeness, and
corresponds with the test for the non-monotonic relationship between population density
and slavery predicted by the unrestricted version of the model. The ordering of columns (1)
through (5) are preserved in Tables 2 through 5.
12These marginal effects are evaluated at the mean, and are calculated using the dprobit command in Stata.
LAND ABUNDANCE 21
The “land abundance” narrative has significant power to explain Egba land tenure. In
1877, an Anglican missionary reported that cultivators could acquire land they developed
from forest, either for free or in return for token payments (Agiri, 1974, p. 467). This could
be true even when land was acquired for planting cocoa or kola. Because land was so cheaply
available, the market for land was thin. European visitors did not believe that the Egba
sold land during the nineteenth century, and even after sales had come into existence many
disputants in the court records stated that they did not believe these to be legitimate. The
use of long fallows – sometimes up to twenty years – economized on labor. Rights over land
were often held only so long as the land was under cultivation, and the “caretakers” left
behind to keep track of a fallow plot could, over time, acquire de facto ownership.
How, then, do we explain land sales and land disputes among the Egba? First, land was
not abundant at all times in Egba society. Mabogunje (1961) notes that during the initial
scramble for land at Abeokuta, township chiefs were required to give up their rights to land
so that newcomers could settle, so that the town would grow larger and more secure from
attack. This devolved control of land to families. Mabogunje (1961) believes that this set
the stage for later land sales. A Boserupian interpretation of his argument would, within
the model, represent this as a shift from SE to SS. Using legendary accounts of the Egba
homeland and travelers’ estimates of Abeokuta’s population during the 1850s and 1860s, it
is clear that the Egba lost over 80% of their territory, and were at least twice as densely
settled in 1830 as they were in 1914. In the area immediately around Abeokuta, fallows were
shorter, intercropping more intense, and forest less present as late as 1902.13
Second, Austin (2008a) has noted that, even while land is abundant, “good” land is
always scarce. For the Egba, lands closer to their settlements and under the protection of
powerful chiefs were more valuable and often the subject of dispute. Within the sample of
court records, land that was more valuable due to cocoa or palm trees was more vigilantly
defended and more likely to be involved in a commercial transaction. Plots endowed with
palm trees were pawned more often, and more frequently defended with the placement of
a caretaker. Cocoa and palm trees both led disputes to have been discussed before the
township chiefs prior to a case coming to court. This could be because disputes were more
common over these plots, or because claimants were willing to expend more effort. Greater
damages were claimed in cases involving cocoa.
5.2. Slavery. Table 3 gives results for slavery. The model’s predictions for slavery confirm
the restricted version of the model. While the point estimate on Ai is positive when only
a North/Sahara dummy is included, it is not significant. Once other controls and regional
dummies are added, this effect grows in magnitude and becomes significant, supporting the
13National Archives of the United Kingdom (NAUK), CO 147/162: 20 Oct, 1902: Acting Governor toChamberlain.
22 JAMES FENSKE
Table 3. Tests of the model: Slavery
(1) (2) (3) (4) (5)
Any slavery
Agric. Suitability 0.03 1.93*** 1.63* 1.66* 1.74**(0.600) (0.715) (0.888) (0.877) (0.815)
Ln(Pop. Density) 0.47**(0.187)
Pop. Density (1960) 1.98(1.295)
Pop. Density Sqd. -0.63(0.456)
Observations 454 454 366 365 366Other Controls No Yes Yes Yes YesRegion F.E. North/Sahara No Yes Yes Yes
Notes: ***Significant at 1%, **Significant at 5%, *Significant at 10%. See notes for Table 2
predictions of the model. This suggests that agricultural suitability is correlated with both
observed and unobserved features that make slavery less likely, and so the estimated impact
is not due solely to omitted variables bias. In Columns (2) and (3), a one standard deviation
increase in agricultural suitability raises the probability of slavery between 1% and 3%.
While the log of population density is positively correlated with the existence of slavery,
there is no significant quadratic correlation of slavery with population density.14 This need
not imply a rejection of the model, for two reasons. First, the disease environment in Africa
may be so severe, and q so high, that the zero-population growth locus LF (Pt) is too steep
to intersect the free labor region, SF . Population simply cannot grow to the point where
free labor replaces slavery. Second, Africa is sparsely populated. There may not be enough
densely-populated societies in the data with relatively low agricultural suitability to identify
the relationship statistically. Both of these fit well with the land abundance view of African
history.
For the Egba, the abundance of land prevented the emergence of wage labor. Even during
the slack season, individuals could gather forest products for themselves. Examples of paid
work in the nineteenth century almost always involve missionaries hiring (or struggling to
hire) laborers. Slavery was, as Oroge (1971) has described, an important means used by
the war chiefs and major traders to secure access to labor where wage work was absent.
Various estimates suggest that slaves were anywhere from one fifth to a “very considerable”
14I have also tested whether population splines or quantiles reveal a significant inverted-U pattern in groupsof 3, 5 and 10. They do not, and these results are not reported.
LAND ABUNDANCE 23
Table 4. Tests of the model: Polygyny
(1) (2) (3) (4) (5)
Any polygyny
Agric. Suitability -0.68 4.69*** 6.06*** 6.06** 5.26*(0.447) (1.672) (2.286) (2.605) (2.858)
Ln(Pop. Density) 0.38(0.413)
Pop. Density (1960) 2.38(3.151)
Pop. Density Sqd. -0.80(2.234)
Observations 517 434 205 203 205Other Controls No Yes Yes Yes YesRegion F.E. North/Sahara No Yes Yes Yes
Notes: ***Significant at 1%, **Significant at 5%, *Significant at 10%. See notes for Table 2
proportion of the population.15 The war chiefs, who in the model had the smallest r, were
the biggest holders of slaves. They were owed captives taken by their soldiers in raids, and
could use their slaves in a variety of other tasks. Most slaves were used where the model
would predict – where A was highest. Male and female slaves were used as porters and
canoe pullers, and female slaves were used in palm oil production. Burton (1863, p. 301)
believed that commerce raised the demand for slaves. British officials and traders, believing
that slavery was indispensable, were afraid to upset the institution. Instead, they moved
to abolish slave dealing (as opposed to slave holding), and worked only to check the worst
abuses by slave owners.
5.3. Polygyny. Table 4 presents results for polygyny. As in the extension in Section 3.3,
the existence of polygyny are more likely in locations where agricultural suitability is high.
86 observations are lost in column (2), since all societies cultivating roots and tubers are
polygynous. Polygyny is positively correlated with population density, though this is not
significant. The marginal effects of either variable are negligible, which is not surprising
given the lack of variation in the dependent variable. This positive correlation runs contrary
to the conventional arguments in the literature on African history, particularly those of
Goody (1969) and Tambiah and Goody (1973). Inequality between men is a precondition for
polygyny, and so the positive correlation between class stratification and population density
suggests that polygyny is not possible in the most egalitarian, sparsely-settled societies.
15See, for example, Oroge (1971, p. 166), Bowen (1857, p. 320), Burton (1863, p. 299) or NAUK, CO 147/133,enc in 4 June, 1898: Denton to Chamberlain, Evidence for 18th day.
24 JAMES FENSKE
Table 5. Land abundance and state stratification
(1) (2) (3) (4) (5)
State stratification
Agric. Suitability 0.93* 1.32** 0.53 0.59 0.08(0.536) (0.665) (0.715) (0.737) (0.746)
Ln(Pop. Density) 0.32**(0.141)
Pop. Density (1960) 2.67*(1.373)
Pop. Density Sqd. -1.43(0.950)
Observations 475 475 475 472 475Other Controls No Yes Yes Yes YesRegion F.E. North/Sahara No Yes Yes Yes
Notes: ***Significant at 1%, **Significant at 5%, *Significant at 10%. See notes for Table 2
Polygyny was one of the strategies Egba farmers used to cope with the shortage of labor.
Bride price, paid to the wife’s parents, involved work, payment of crops, assistance with
major expenses, and transfers of cash. Coercion and violence could be used to keep a women
married (Byfield, 1996). Payment of bride price established claims over children. Junior
wives worked for senior wives, and all wives worked for their husbands’ other male relatives.
Women did the bulk of “domestic” labor and processed palm fruits into oil. The largest
harems, however, were those of the war chiefs, who could have up to two hundred wives at
the peak of their power. Inequality spurred polygyny in Egba society. Outright purchase of
slaves as wives was common during the nineteenth century.
5.4. State Power. In Table 5, I report results for state stratification. Surprisingly, it is
not related to Ai once regional fixed effects are included. Validating Herbst (2000), it is
correlated with population density, but intrinsic agricultural productivity does not appear
to be one of its systematic determinants. This may reflect the prevalence of African states,
such as ancient Ghana, which drew their revenues from access to trade routes and mineral
resources in environments that were relatively hostile to agriculture. Also surprisingly, the
(insignificant) marginal effects are relatively large – a one standard deviation increase in
agricultural suitability prompts a 3% to 8% increase in the probability of state centralization,
depending on the column. For population density, a one standard deviation rise in log
population is associated with a 17% increase in the chance of a centralized state.
Fitting with the “land abundance” narrative, authority in pre-colonial Egba society was
decentralized, and chiefs did not derive their revenues from land. Political power lay at the
township level, and was divided among the the olorogun (war chiefs), ogboni (civil chiefs), ode
LAND ABUNDANCE 25
Table 6. Neighbor effects
(1) (2) (3) (4)
Any
individual
land rights Any slavery Any polygyny
State
stratification
Spatial lag (ρ) -0.22 0.70 0.23 0.3190% CI [-0.36,-0.06] [0.31,0.96] [-0.05,0.63] [0.16,0.44]
Agric. Suitability 2.16 2.09 -2.01 0.5990% CI [0.48,3.91] [0.37,4.19] [-4.48,0.26] [-0.23,1.46]
Region FE North/Sahara North/Sahara North/Sahara North/SaharaOther Cont. Yes Yes Yes YesObservations 368 454 434 475
Notes: 90% confidence intervals in brackets. All regressions are spatial probit, with coefficients reported(see LeSage and Pace (2009)). See notes for Table 2
(hunters), and parakoyi (trade chiefs). Before Abeokuta, there was no political organization
above the level of the township. While the Alake (king) may have had judicial supremacy
over the Ake townships, he had no authority over the other “provinces” of Egba society.
Political loyalties in Abeokuta after 1830 continued center on townships (Pallinder-Law,
1973, p. 17). Political leaders at Abeokuta derived their powers from military strength. The
most important political leader among the Egba when they arrived Abeokuta was Sodeke,
the Seriki. “Seriki,” or “general of the youths” (Biobaku, 1952) was a military title. His
successor, Okukenu, was eventually given the restored title of Alake, but was described as
“too weak to assert himself when challenged” (Pallinder-Law, 1973, p. 28). Egba political
history to 1893, when the British intervened to create a united central government, is of
competing interests limiting the power of the nominally most senior authorities.
In addition, the Egba did not prosecute war to capture land. Rather, three concerns
drove Egba military adventures – trade, captives, and security. Pallinder-Law (1973, p. 19)
points to access to the coast as the Egba motivation for capturing Ota. When the town was
defeated, the inhabitants were allowed to remain so long as they did not rebuild their walls
(Losi, 1924, p. 56). Losi (1924, p. 87) stresses that the Egba sacked Ikorodu in 1865 to gain
access to the Lagos lagoon and as punishment for supporting their enemies during the Kutuye
war. Ajisafe (1924, p. 65-67) notes that after defeating several townships of Ijebu Remo,
“as was the custom, they left without taking absolute possession of the towns”; similarly,
when they defeated the Egbado in retaliation for “harassment” at Oke Ogun, “they did not
take possession.” Descriptions of Egba wars during the period are replete with mention that
captives were brought back as slaves, whether in the 1862 Makun war (Champness, 1907,
p. 113), or the descriptions given by Losi (1924) of the Dado (1834) and Iperu (1835) Wars.
26 JAMES FENSKE
5.5. Institutional spillovers. The estimate of ρ for land rights is negative. A society with
land rights discourages its neighbors from having rights over land. In the model, if neighbors
switch from egalitarianism to slavery, creating rights over land in their own societies, this
will increase R. This raises the population threshold above which free labor is preferred to
slavery, because more people are drawn out each period. An explanation from outside the
model is that, if secure rights or other “good” institutions correlated with these encourage in-
migration, this will depopulate neighboring regions, delaying the development of land rights.
Positive spill-overs exist in slavery; if a society uses slaves it encourages its neighbors to do
the same. This may be because it is more difficult for a slave to escape to a neighboring slave
society, lowering r. It may also, outside the model, reflect the military value of slaves and
the need for a society to defend itself from neighboring militarized societies. The impact of
agricultural suitability disappears for polygyny in the spatial regressions because these results
are sensitive to the weighting of observations; this is not surprising, given the lack of variation
in these outcomes. There is no evidence that marital institutions show correlation across
space. Finally, state stratification displays positive neighbor effects. This may reflect the
need for societies to defend themselves against their organized neighbors. This may also be
due to direct institutional spillovers. Oral tradition, for example, states that the institution
of kingship was transferred directly from Ife to Benin during the thirteenth century (Ryder,
1965). Similarly, the formation of the Swazi and Lesotho states was a direct response to the
rise of Zulu power during the mfecane (Maddox, 2006, p. 114).
For the Egba, institutional spillovers in land tenure run contrary to those suggested by
the econometric results. This was because the Egba had Lagos as a southern neighbor.
After 1861, this was a British colony. It was through Lagos that missionaries and mission-
educated repatriated slaves came to Lagos, introducing ideas of individual ownership, and
asking to purchase land in freehold as they had in Sierra Leone (Mabogunje, 1961). The
Egba also influenced land tenure in Lagos. After an anti-Christian uprising in 1867, many
Egba converts fled to Lagos, and were allotted parcels by the Governor on land given to him
by a Lagos chief. Over time, these came to be viewed largely as freehold grants and were
one of the spearheads for alienability of land in Lagos (Mann, 2007). In the case of slavery,
by contrast, the Egba gained from their neighbors’ practices; by mid-century, slaves were
increasingly purchased in markets to the North, in Rabba and Ilorin. By 1870, “Hausa”
slaves were the majority in Abeokuta (Agiri, 1981, p. 137). These northerners were far from
home and less likely to flee. Anti-slavery policies in Lagos gave Egba slaves a means of
escape, and led to political crises between the two states (Oroge, 1975). What efforts did
exist to form a centralized authority in Abeokuta was motivated by military concerns – what
strength existed in the Egba state was necessitated by the rise of rivals such as Ibadan and
Ijebu (Ajayi and Smith, 1964).
LAND ABUNDANCE 27
In sum, the model correctly predicts that land rights and polygyny existed in pre-colonial
Africa where population and agricultural suitability were greatest. As in the model, slavery
existed where agricultural suitability was high, but population was positively correlated with
slavery. This is consistent with the literature’s characterization of Africa as land abundant.
Due to geographic factors and the disease environment, population could not expand in the
African context to the point where free labor would replace slavery. While state power was
correlated with population density, its existence was not systematically based on agricultural
suitability. The nature of Egba land tenure, slavery, polygyny, and state power can all be
understood within the “land abundance” narrative that is formalized by the model in Section
3.
6. Robustness
In this section, I show that the results in Section 5 are not solely due to the use of
dependent variables with little variation in their outcomes, that they can be replicated with
an alternative measure of historical population density, that they are robust to additional
controls that represent alternative interpretations of the results, and that they are better
explained by the model than by other theories of slavery.
6.1. Dependent variables. Because there is not much variation in the existence of land
rights and polygyny in the data, I use alternative measures of each. The first is an indicator
for whether the inheritance of land is patrilineal.16 Following Goody (1969), this captures
the degree to which the control of real property is directed towards the nuclear family.
Roughly, this is one step along the transition from weakly defined to strongly defined rights
in land. Similarly, there is not much variation in the presence of polygyny. Hence, the
“willingness to pay” for wives is measured by using an indicator for whether consideration
is given in return for a bride (a non-token bride-price, labor service, or another female
relative).17 Sadly, there is not much variation in this alternative measure either. Finally,
measuring state stratification as the number of levels of jurisdiction may not capture the
existence of stratified, albeit localized states. Hence, class stratification among freemen is
used as an alternative measure of the presence of states.18 Summary statistics for these
variables are given in Table 1.
Table 7 replicates columns (3) and (4) from each of Tables 2, 4 and 5 using these alterna-
tive measures. The results generally follow the same pattern as in Section 5. Patrilineal land
inheritance is positively correlated with both agricultural suitability and population density,
16Like the indicator for land rights, this is constructed using V74: Inheritance Rule for Real Property (Land).This is equal to 1 if V74=4, V74=5, V74=6, or V74=7.17This is constructed using V6: Mode to Marriage (Primary). Consideration for bride exists if V6=1 orV6=2 or V6=5.18This uses V66: Class Stratification. Class stratification exists if V66>1.
28 JAMES FENSKE
Table 7. Alternative dependent variables
(1) (2) (3) (4) (5) (6)
Land inheritance
is patrilineal
Consideration
given in return
for bride
Class
stratification
Agric. Suitability 1.72* 1.62 3.44* 3.43* -0.33 -0.37(1.033) (1.076) (2.039) (1.988) (1.020) (1.009)
Ln(Pop. Density) 0.33*** 0.13 0.21(0.109) (0.226) (0.164)
Observations 354 354 413 412 401 400Other Controls Yes Yes Yes Yes Yes YesRegion F.E. Yes Yes Yes Yes Yes Yes
Notes: ***Significant at 1%, **Significant at 5%, *Significant at 10%. See notes for Table 2
though the link with agricultural suitability is only marginally significant in Column (2).
Payment of consideration for brides is again positively correlated with agricultural suitabil-
ity, though there is generally only a weak positive relationship with population density. All
societies in the North or Sahara offer more than a token price for wives, which reduces the
sample size. Class stratification, like state stratification, is unrelated to agricultural suitabil-
ity, though the positive correlation with population density is now statistically insignificant.
With a p-value of less than 0.2, this is still somewhat suggestive. In general, while these al-
ternative measures are not as statistically robust as the more direct measures used in Section
5, they yield qualitatively similar results.
6.2. Population density. The use of population density in 1960 is necessitated by data
availability. There are two possible problems with its use. First, population growth between
1900 and 1960 may have been determined by the institutions that prevailed on the eve of
colonial rule. For this to produce spurious correlations, however, would require restrictive and
implausible narratives about the nature of colonial rule. For example, if there were no true
correlation between land rights and population density, societies with rights to land would
have to have grown especially quickly as a result of colonial rule to produce the observed
results. Colonial policies that promoted population growth (or crowded people together)
would have had to have been targeted particularly at societies in which land rights were well
developed. Medical innovations were largely introduced after 1950, and would have come
too late to have an appreciable effect. Similarly, the confinement of Africans onto reserves
in countries such as Kenya was not motivated by their pre-colonial institutions, but instead
by the attractiveness of the areas they inhabited.
Second, population density in 1960 may be a poor historical proxy. The strong correlation
between it and population density in 2000 mentioned above suggests that this concern is
LAND ABUNDANCE 29
Table 8. Population density projected to date of observation
(1) (2) (3) (4)
Any individual land rights Any slavery
Agric. Suitability 3.11*** 2.48*** 1.81** 1.67**(0.958) (0.756) (0.884) (0.836)
Ln(Pop. Den., D.O.) 0.16** 0.35*(0.075) (0.186)
Pop. Density (D.O.) -12.28** 3.53(5.798) (2.613)
Pop. Density Sqd. 75.22*** -3.51(21.344) (2.295)
Observations 321 321 365 365Other Controls Yes Yes Yes YesRegion F.E. Yes Yes Yes Yes
Any polygyny State stratification
Agric. Suitability 6.19** 5.74** 0.60 0.61(2.486) (2.663) (0.705) (0.839)
Ln(Pop. Den., D.O.) 0.29 0.25*(0.367) (0.138)
Pop. Density (D.O.) -20.80* 0.31(10.859) (1.595)
Pop. Density Sqd. 47.00*** -0.03(16.566) (0.947)
Observations 203 203 472 472Other Controls Yes Yes Yes YesRegion F.E. Yes Yes Yes Yes
Notes: ***Significant at 1%, **Significant at 5%, *Significant at 10%. See notes for Table 2
minor. In this section I use the growth rate between 1960 and 2000 to impute an alternative
proxy for historical population density. I project back to the date of observation reported
in the Ethnographic Atlas. While the range of observed growth rates is fairly large, this
is uncorrelated with initial population density, and so is not likely to be systematically
biased.19 These growth rates have been stable over time; population growth from 1960 to
1980 is tightly correlated with growth from 1960 to 2000.20 Table 8 replicates Columns
(4) and (5) of Tables 2, 3, 4 and 5. The positive correlations of all four institutions with
population density are apparent. For polygyny this is still insignificant.
6.3. Other robustness checks. In addition to the lack of variation in the dependent vari-
ables and the validity of the proxy for population density, there are other objections that
19The correlation coefficient is -0.0614, with p-value 0.154.20The correlation coefficient is 0.8234, with p-value 0.000.
30 JAMES FENSKE
may be raised about the approach used to test the model. I have tested the robustness of
the results to these objections in Table 9. Each odd-numbered column replicates Column
(3) from Table 2, 3, 4 or 5 – the principal specification for agricultural suitability. Each
odd-numbered column does the same with column (4) from these earlier tables, reproducing
the main specification for population density. To conserve on space, I have only reported
the p-values on agricultural suitability and the log of population density, to show whether
the estimates remain statistically significant.
First, it is not clear to what degree identification of the results are coming off differences
within regions that are not representative of the whole of Africa. I present the results of the
main specifications excluding the Saharan and North African regions. Second, because of
the possible endogeneity of soil depth and soil fertility, I add variables that control directly
for soil depth and soil fertility constraints.21 Third, while data on religion is not available,
exposure to Islam is likely to directly affect land rights, slavery, polygyny, and states. As a
proxy for this and other North African cultural influences, I add distance from the nearest
of: Cairo, Tripoli, Tunis, Algiers and Casablanca. Third, it is clear from visual inspection
of Figure 2 that agricultural suitability may be systematically mis-measured in areas where
major rivers provide access to water even in the presence of precipitation constraints, or
where rivers can facilitate trade. I thus present the results including a dummy for whether
each ethnic group’s territory is intersected by the Nile, and a dummy for whether any major
river (the Benue, Blue Nile, Chire, Congo, Lualaba, Lukaga, Niger, Nile, Orange, Ubangi,
White Nile, or Zambezi) intersects the group’s territory.
Fourth, an alternative hypothesis is that agricultural suitability and population density
are proxies for urbanization, which is itself responsible for the existence of land rights and
slavery. To control for pre-colonial urbanization, I add the number of cities in 1850 given
by Chandler and Fox (1974), who take 40,000 as the cutoff population for including a city
in their list. Because greater agricultural suitability may also be correlated with greater
pastureland suitability, I add the society’s percentage dependence on animal husbandry.22
Finally, to ensure that identification is not driven by comparison of small, non-cultivating
societies with settled agrarian groups, I have included percentage dependence on agriculture
as an additional control.23
While many of the results look very similar in the baseline and the robustness specifica-
tions, there are three major exceptions. First, including soil fertility constraints makes the
effect of agricultural suitability on land rights disappear. Second, discarding North Africa
or including the Nile renders the relationship between agricultural suitability and slavery
insignificant, while making the effect of agricultural suitability on polygyny negative and
21These are Plates 21 and 22 in the FAO data.22This is V4 in the Ethnographic Atlas.23This is V5 in the Ethnographic Atlas.
LAND ABUNDANCE 31
Table 9. Other Robustness Checks: P-values on A and ln(P )
(1) (2) (3) (4) (5) (6) (7) (8)
Robustness CheckLand
Rights Slavery Polygyny
State
Stratifi-
cation
A ln(P) A ln(P) A ln(P) A ln(P)
Baseline 0.00 0.00 0.07 0.01 0.01 0.36 0.45 0.02Drop Sahara 0.00 0.00 0.07 0.01 0.02 0.45 0.54 0.03Drop North Africa 0.00 0.00 0.39 0.03 0.00∗ 0.00 0.61 0.26Add Dist. to Islamic Cities 0.01 0.00 0.09 0.02 0.08 0.62 0.51 0.00Add Soil Depth 0.00 0.00 0.06 0.02 0.00 0.34 0.72 0.02Add Soil Fertility 0.74 0.05 0.02 0.00 0.00 0.15 0.88 0.03Add Nile 0.00 0.00 0.16 0.02 0.01 0.39 0.31 0.03Add Any Major River 0.00 0.00 0.06 0.01 0.01 0.00 0.53 0.01Add Cities in 1850 0.00 0.00 0.12 0.08 0.35 0.83 0.76 0.44Add Husbandry 0.00 0.00 0.05 0.01 0.00 0.00 0.49 0.03Add Agriculture 0.00 0.02 0.05 0.00 0.00 0.08 0.61 0.05
Notes: Odd-numbered columns include other controls and regional fixed effects, while even-numberedcolumns additionally control for agricultural suitability. See notes for Table 2. Soil depth is the FAO-GAEZ measure of constraints on soil depth. Soil fertility is the FAO-GAEZ measure of constraints onsoil fertility. Nile is a dummy for whether the ethnic group is intersected by the Nile. Any major riveris a dummy for whether the ethnic group is intersected by the Benue, Blue Nile, Chire, Congo, Lualaba,Lukaga, Niger, Nile, Orange, Ubangi, White Nile, or Zambezi. Cities in 1850 is the number of cities listed byChandler and Fox (1974) in the ethnic group’s territory (these are cities of 40,000 and above). Husbandry isthe percentage dependence of the society on animal husbandry (V4 in the Ethnographic Atlas). ∗ indicatesthat the effect, while statistically significant, is negative.
significant. Third, the correlation between state stratification and population density is not
robust to the exclusion of North Africa or the inclusion of pre-colonial urbanization.
Soil fertility constraints enter strongly and negatively into the land rights equation. While
it is hard to imagine soil fertility predicting land rights except through the suitability of
the land for agriculture, this suggests that one of the most important dimensions through
which agricultural suitability affects land rights is soil fertility. Insofar as more well defined
rights over land create incentives to preserve soil fertility, this suggests that the parameter
estimates in the baseline are overstated due to endogeneity. To believe that the entire result
is spurious, however, would require that the entire correlation is due to reverse causation.
This would contradict studies that have found the relationship between land tenure and soil
conservation investments in Africa to be surprisingly weak – see for example Besley (1995)
or Hagos and Holden (2006).
Sensitivity of the slavery result to the exclusion of North Africa or addition of the Nile is
more problematic. That the polygyny results are sensitive is not surprising, given their lack
32 JAMES FENSKE
of variability. This result may be in part due to greater variance in the estimates; the causal
effect of agricultural suitability is still significant when regional fixed effects are excluded. As
well, with a standard deviation of 0.24, North Africa has the most within-region variability
in agricultural suitability that can be used for identification. Without region fixed effects,
if an interaction term is included between North Africa and agricultural suitability, both
the direct effect and interaction are significant. What this suggests is that the impact of
agricultural suitability is present both in North Africa and Sub-Saharan Africa, but that the
effect is stronger in the North and along the lower Nile. This may be due to the greater
ability to import technologies that intensify agriculture where population density is high and
the land is suitable.
That states are not correlated with population in the absence of North Africa is surprising.
If regional fixed effects are excluded, the correlation is still significant, which suggests that
this may result in part from multi-collinearity and the loss of observations. That cities
eliminate the relationship between population density and states is not as surprising. Cities
facilitate centralized administration and are, of course, highly correlated with population
density.
6.4. Theories of slavery. In this section, I contrast the results outlined in Section 5 with
two other major theories of slavery and explain why the model outlined in Section 3 does a
better job of explaining African slavery than either of these.
First, writers such as Inikori (1999) have suggested that African “slaves” held a position
closer to that of the European serf. In the model, slaves are coerced workers whose price
does not depend on the local supply of labor. The severity of slavery is not important to
this conceptual distinction. The dominant theory of serfdom is that of North and Thomas
(1971), who hold that serfs voluntarily exchanged their labor for protection from lords. These
payments were in inputs rather than money because of the limited nature of output markets.
There are at least four reasons why this model cannot explain Africa. First, that model’s
applicability to any case has been called into question by Fenoaltea (1975), who demon-
strates that North and Thomas (1971) err in treating serfdom as voluntary, underestimate
the transactions costs in labor contracts, misidentify the historical trends that acted on the
manorial system, and overemphasize the rigidity of “custom” in constraining institutional
change. Second, both agricultural suitability and population density have been shown in
Section 5 to be positively associated with slavery. In the North and Thomas (1971) model,
these should promote the development of trade and markets, lessening the need for con-
tracts to be written labor dues. Third, their model predicts that trade will discourage the
use of serfs. This runs counter to the literature on African history, which has shown that
external trade in particular spurred greater use of slaves in production (e.g. Lovejoy (2000)
or Law (1995)). Finally, there is no evidence that African slaves received payments that
LAND ABUNDANCE 33
approximated their marginal products. In many cases, slaveowners had to be compelled to
receive manumission payments from their slaves under colonial rule, suggesting that they
were earning rents for which they needed to be compensated. This was true for the Egba,
where colonial courts took note of manumission payments into the 1920s. Austin (2009b),
similarly, provides several examples from nineteenth century West Africa in which it was
possible for the purchaser of a slave to recoup his investment within six years.
The second theory of slavery I address is the collection of arguments that, in certain
contexts, slavery is more productive than free labor, which explains its use. For Fenoaltea
(1984), this occurs where “pain incentives” are effective and detailed care by the worker is
unnecessary. Fogel and Engerman (1974) link the exceptional productivity of slaves in the
American south to economies of scale that could only be achieved through gang labor, an
activity so grueling that free men could not be induced to take part at any price. Engerman
and Sokoloff (1997), similarly, argue that the cultivation of crops with economies of scale is
more conducive to slavery. Hanes (1996) explains the concentration of slaves in rural and
domestic production by invoking the high turnover costs in these industries.
These arguments again cannot alone explain slavery in Africa. First, there is no evidence
that slaves were used in production in sectors systematically different than those dominated
by free peasants. The fact that, over a few generations, slaves were often partly assimilated
into their masters’ societies is evidence that they were not kept in economic isolation (Austin,
2009b). Where large slave communities of slaves were present, (see e.g. Lovejoy (1978) for
the Sokoto Caliphate or Oroge (1971) for nineteenth century Yorubaland), these existed
not because slaves were used in economic tasks that free peasants were not, but because
they were acquired in large numbers by authorities and other elites. Studies of slavery in
individual African societies frequently make reference to slave labor and free labor working
in the same tasks. Austin (2005) notes gold and kola production in Asante were both carried
out by free people, pawns, corvee labor, slaves, and descendants of slaves. Uchendu (1979)
shows for Igbo society that slaves first were used to fill subsistence needs by farming and
fishing, and only secondarily filled prestige functions. “In domestic activities,” he argues,
“no operation was strictly reserved for slaves.” Describing the Kerebe of Tanzania, Hartwig
(1979) writes that masters often worked alongside their slaves, who performed the same tasks
as their owners and their owners’ wives.
Second, the literature on the “legitimate commerce” period suggests that slaves were
used in the activities where labor of all kinds was most productive; in the model this is
consistent with a rise in A, and does not require a different production function under slavery.
The nineteenth century export markets for oils, ivory, ostrich feathers and other goods
created higher returns to slave labor, and slavery within Africa intensified (Lovejoy, 2000).24
24Lynn (1997) also provides a survey of the period, while Law (1995) contains a number of case studies.
34 JAMES FENSKE
Third, African agriculture both past and present has been overwhelmingly characterized
by diminishing or constant returns to scale (Hopkins, 1973). Without evidence of scale
economies, an appeal to “pain incentives” is not necessary to explain slavery over and above
a comparison of the costs of slavery to those of free labor.25
7. Conclusion
Bad institutions are one of the fundamental causes of African poverty, and the institu-
tions that exist on the continent currently have been shaped by those that existed prior to
colonial rule. I have addressed a theme in the economics literature – how geography affects
institutions – by looking in depth at one hypothesis from the literature on African history. I
find that African land tenure, slavery, polygyny, and states have all been decisively shaped
by the continent’s abundance of land and scarcity of labor. I find that this perspective ex-
plains much about institutions in pre-colonial Africa, using cross-sectional evidence. That
the Egba can be so easily understood in terms of the “land abundance” view gives that
narrative further support.
The use of a formal model and comparative data have made several points that must be
taken into account in understanding the impacts of under-population on African institutions.
First, when both productivity and population are low, the opportunity cost of coercion is
high, and the benefit to creating estates is low. This explains why slavery is less common
among the most sparsely populated African societies. Second, greater agricultural suitability
(as well as access to trade), will encourage increased reliance on slavery. This explains why
some of the most agriculturally prosperous though densely populated regions in Africa, such
as Sokoto, also used slaves most intensively (cf. Hill (1985)). Third, where brides were
costly and polygyny existed in pre-colonial Africa, agricultural productivity (and hence the
marginal product of labor) was highest, but population density was also greater. Inequality,
then, is a prerequisite for unequal access to wives. Fourth, state strength in Africa has
been associated with population density, but is not systematically related to agricultural
productivity. Finally, there are substantial institutional spillovers across African societies
relating to land, slavery, and the power of states. These revisions to the current thinking
allow the “land abundance” perspective to better explain institutions and are borne out in
comparative data.
25Returning to the model, if slaves are worked harder than free laborers, their productivity may be enhancedby some factor η. This parameter will carry over into the definitions of Φ, Ψ, and Ω. However, unless theshape of the production function itself changes, the qualitative shapes of the institutional regions will notbe different.
LAND ABUNDANCE 35
References
Acemoglu, D., Cantoni, D., Johnson, S., and Robinson, J. A. (2009). The consequences of
radical reform: The French Revolution. Working Paper.
Acemoglu, D., Johnson, S., and Robinson, J. (2005). The rise of Europe: Atlantic trade,
institutional change, and economic growth. American Economic Review, 95(3):546–579.
Acemoglu, D., Johnson, S., and Robinson, J. A. (2001). The colonial origins of comparative
development: An empirical investigation. The American Economic Review, 91(5):1369–
1401.
Acemoglu, D., Johnson, S., and Robinson, J. A. (2003). An African Success Story: Botswana.
In Search of Prosperity: Analytical Narratives on Economic Growth. Princeton University
Press, Princeton, NJ.
Adjuik, M., Bagayoko, M., Binka, F., Coetzee, M., Cox, J., Craig, M., Deichman, U.,
deSavigny, F. E. D., Fraser, C., and Gouws, E. (1998). Towards an atlas of malaria
risk in Africa: First technical report of the Mapping Malaria Risk in Africa collaboration.
MARA/ARMA, Durban.
Agiri, B. (1974). Aspects of socio-economic changes among the Awori Egba and Ijebu Remo
communities during the nineteenth century. Journal of the Historical Society of Nigeria,
7(3):465–483.
Agiri, B. (1981). Slavery in Yoruba Society in the 19th century, pages 123–148. The Ideology
of Slavery in Africa. Sage Publications, Beverly Hills.
Ajayi, J. F. A. and Smith, R. (1964). Yoruba warfare in the nineteenth century.
Ajisafe, A. K. (1964 [1924]). History of Abeokuta. Fola Bookshops, Abeokuta.
Altonji, J., Elder, T., and Taber, C. (2005). Selection on observable and unobservable
variables: Assessing the effectiveness of Catholic schools. Journal of Political Economy,
113(1):151–184.
Austin, G. (2004a). Markets with, without, and in spite of states: West Africa in the
pre-colonial nineteenth century. Working Paper.
Austin, G. (2004b). Sub-Saharan Africa: land rights and ethno-national consciousness in
historically land abundant economies, pages 276–293. Land Rights, Ethno-nationality and
Sovereignty in History. Routledge, London.
Austin, G. (2005). Labour, land, and capital in Ghana: from slavery to free labour in Asante,
1807-1956. University of Rochester Press, Rochester, NY.
Austin, G. (2008a). Resources, techniques, and strategies south of the Sahara: revising
the factor endowments perspective on African economic development, 1500-2001. The
Economic History Review, 61(3):587–624.
36 JAMES FENSKE
Austin, G. (2008b). The “reversal of fortune” thesis and the compression of history: Per-
spectives from African and comparative economic history. Journal of International Devel-
opment, 20(8):996–1027.
Austin, G. (2009a). Coercion and markets: Integrating economic and social explanations of
slavery in precolonial West Africa, c. 1450-1900. Paper Presented at the World Economic
History Congress, Utrecht.
Austin, G. (2009b). Factor markets in Nieboer conditions: Pre-colonial West Africa, c.
1500-c. 1900. Continuity and Change, 24(01):23–53.
Banerjee, A. and Iyer, L. (2005). History, institutions, and economic performance: The
legacy of colonial land tenure systems in India. American Economic Review, 95(4):1190–
1213.
Berry, S. (1988). Property rights and rural resource management: The case of tree crops in
West Africa. Cahier des sciences humaines, 24(1):3–16.
Bertocchi, G. and Canova, F. (2002). Did colonization matter for growth? An empirical
exploration into the historical causes of Africa’s underdevelopment. European Economic
Review, 46(10):1851–1871.
Besley, T. (1995). Property rights and investment incentives: Theory and evidence from
Ghana. Journal of Political Economy, 103(5):903–37.
Biobaku, S. O. (1952). An historical sketch of Egba traditional authorities. Africa, 22(1):35–
49.
Bloom, D. E. and Sachs, J. D. (1998). Geography, demography, and economic growth in
Africa. Brookings Papers on Economic Activity, 1998(2):207–295.
Bowen, T. J. (1857). Central Africa: Adventures and missionary labors in several countries
in the interior of Africa, from 1849 to 1856. Sheldon, Blakeman, New York.
Bubb, R. (2009). States, law, and property rights in West Africa. Working Paper.
Burton, R. F. (1863). Abeokuta and the Cameroons Mountains: An exploration. Tinsley
Brothers, London.
Byfield, J. (1996). Women, marriage, divorce and the emerging colonial state in Abeokuta
(Nigeria) 1892-1904. Canadian Journal of African Studies, 30(1):32–51.
Champness, E. M. (1907). The life-story of Thomas Champness. C. H. Kelly.
Chandler, T. and Fox, G. (1974). 3000 years of urban growth. Academic Press.
Collier, P. and Gunning, J. W. (1999a). Explaining African economic performance. Journal
of Economic Literature, 37(1):64–111.
Collier, P. and Gunning, J. W. (1999b). Why has Africa grown slowly? The Journal of
Economic Perspectives, 13(3):3–22.
Dell, M. (2009). The persistent effects of Peru’s mining Mita. Working Paper.
Dowd, J. (1917). Slavery and the slave trade in Africa. Journal of Negro History, 2(1):1–20.
LAND ABUNDANCE 37
Easterly, W. and Levine, R. (2003). Tropics, germs, and crops: How endowments influence
economic development. Journal of Monetary Economics, 50(1):3–39.
Engerman, S. L. and Sokoloff, K. L. (1997). Factor endowments, institutions, and differential
paths of growth among new world economies, pages 260–304. How Latin America Fell
Behind. Essays on the Economic Histories of Brazil and Mexico, 1800-1914. Stanford
University Press.
Fage, J. D. (1969). Slavery and the slave trade in the context of West African history.
Journal of African History, 10(3):393–404.
Fenoaltea, S. (1975). The rise and fall of a theoretical model: The manorial system. Journal
of Economic History, 35(2):386–409.
Fenoaltea, S. (1984). Slavery and supervision in comparative perspective: A model. The
Journal of Economic History, 44(3):635–668.
Fischer, G., Shah, M., Velthuizen, H., and Nachtergaele, F. O. (2002). Global Agro-Ecological
Assessment for Agriculture in the 21st Century: Methodology and results. International
Institute for Applied Systems Analysis/Food and Agriculture Organization of the United
Nations, Laxenburg and Rome.
Fogel, R. W. and Engerman, S. L. (1974). Time on the cross: The economics of American
Negro slavery. Little, Brown and Company, Boston.
Gallup, J. L. and Sachs, J. D. (2001). The economic burden of malaria. The American
Journal of Tropical Medicine and Hygiene, 64(1,2):85–96.
Gennaioli, N. and Rainer, I. (2007). The modern impact of precolonial centralization in
Africa. Journal of Economic Growth, 12(3):185–234.
Goody, J. (1969). Inheritance, property, and marriage in Africa and Eurasia. Sociology,
3(1):55–76.
Goody, J. (1980). Slavery in time and space, pages 16–42. Asian and African Systems of
Slavery. University of California Press, Berkeley and Los Angeles.
Gordon, R. G. and Grimes, B. F. (2005). Ethnologue: Languages of the world. SIL Interna-
tional, Dallas, TX.
Greif, A. (2006). Institutions and the path to the modern economy: Lessons from medieval
trade. Cambridge University Press, Cambridge.
Grindal, B. T. (1972). An ethnographic classification of the Sisala of northern Ghana.
Ethnology, 11(4):409–424.
Hagos, F. and Holden, S. (2006). Tenure security, resource poverty, public programs, and
household plot-level conservation investments in the highlands of northern ethiopia. Agri-
cultural Economics, 34(2):183–196.
Hanes, C. (1996). Turnover cost and the distribution of slave labor in Anglo-America. The
Journal of Economic History, 56(2):307–329.
38 JAMES FENSKE
Hartwig, G. W. (1979). Changing forms of servitude among the Kerebe of Tanzania, pages
261–287. Slavery in Africa: historical and anthropological perspectives. University of
Wisconsin Press.
Herbst, J. (2000). States and Power in Africa: Comparative Lessons in Authority and
Control. Princeton University Press.
Hill, P. (1985). Comparative West African Farm-slavery Systems, volume 2 of Slaves and
Slavery in Muslim Africa, pages 33–50. Frank Cass, London.
Hopkins, A. G. (1973). An economic history of West Africa. Columbia University Press,
New York.
Iliffe, J. (1995). Africans: the history of a continent. Cambridge University Press, Cambridge.
Inikori, J. E. (1999). Slaves or Serfs? A Comparative Study of Slavery and Serfdom in
Europe and Africa, pages 49–75. The African Diaspora: African Origins and New World
Identities. Indiana Univ Press.
Isaacman, A., Stephen, M., Adam, Y., Homen, M. J., Macamo, E., and Pililao, A. (1980).
“Cotton is the mother of poverty”: Peasant resistance to forced cotton production in
Mozambique, 1938-1961. International Journal of African Historical Studies, 13(4):581–
615.
Kjekshus, H. (1977). Ecology control and economic development in East African history.
University of California Press.
Kopytoff, I. (1987). The internal African frontier: The making of African political culture,
pages 3–84. The African frontier: The reproduction of traditional African societies. Indiana
University Press.
Kopytoff, I. and Miers, S. (1977). African ‘slavery’ as an institution of marginality. Slavery
in Africa: historical and anthropological perspectives, pages 3–81.
Lagerlof, N. P. (2009). Slavery and other property rights. Review of Economic Studies,
76(1):319–342.
Law, R. (1995). From Slave Trade to “Legitimate” Commerce. Cambridge University Press,
Cambridge.
LeSage, J. P. and Pace, R. K. (2009). Introduction to spatial econometrics. CRC Press.
Losi, J. B. O. (1924). History of Abeokuta. Bosere Press, Lagos, Nigeria.
Lovejoy, P. E. (1978). Plantations in the economy of the Sokoto Caliphate. Journal of
African History, pages 341–368.
Lovejoy, P. E. (2000). Transformations in Slavery: A History of Slavery in Africa. Cambridge
University Press.
Lynn, M. (1997). Commerce and economic change in West Africa: The palm oil trade in the
nineteenth century. Cambridge University Press, Cambridge.
LAND ABUNDANCE 39
Mabogunje, A. L. (1961). Some comments on land tenure in Egba Division, western Nigeria.
Africa: Journal of the International African Institute, 31(3):258–269.
Maddox, G. (2006). Sub-Saharan Africa: An Environmental History. Abc-Clio Inc.
Mahadi, A. and Inikori, J. E. (1987). Population and capitalist development in precolonial
West Africa: Kasar Kano in the nineteenth century, pages 62–73. African Population
and Capitalism: Historical Perspectives. The University of Wisconsin Press, Madison,
Wisconsin.
Mann, K. (2007). Slavery and the birth of an african city: Lagos, 1760-1900.
McCann, J. (1999). Green Land, Brown Land, Black Land: An Environmental History of
Africa, 1800-1990. Heinemann.
Miers, S. and Klein, M. A. (1998). Introduction. Slavery & Abolition, 19(2):1–15.
Murdock, G. P. (1959). Africa: Its Peoples and Their Culture History. McGraw-Hill.
Murdock, G. P. (1967). Ethnographic atlas. University of Pittsburgh Press.
Nelson, A. and Deichmann, U. (2004). African Population Database Version 4. United
Nations Environment Programme, Sioux Falls, SD, USA.
North, D. C. (1991). Institutions. The Journal of Economic Perspectives, 5(1):97–112.
North, D. C. and Thomas, R. P. (1971). The rise and fall of the manorial system: A
theoretical model. The Journal of Economic History, 31(4):777–803.
North, D. C. and Weingast, B. R. (1989). Constitutions and commitment: The evolution of
institutional governing public choice in seventeenth-century England. Journal of Economic
History, 49(4):803–832.
Northrup, D. (1979). Nineteenth-century patterns of slavery and economic growth in south-
eastern Nigeria. International Journal of African Historical Studies, 12(1):1–16.
Nunn, N. (2008). The long-term effects of Africa’s slave trades. The Quarterly Journal of
Economics, 123(1):139–176.
Nunn, N. (2009). The importance of history for economic development. Annual Review of
Economics, 1(1):1–28.
Nunn, N. and Puga, D. (2007). Ruggedness: The blessing of bad geography in Africa.
Working Paper.
Oroge, E. A. (1971). The institution of slavery in Yorubaland with particular reference to
the nineteenth century.
Oroge, E. A. (1975). The fugitive slave question in Anglo-Egba relations 1861-1886. Journal
of the Historical Society of Nigeria, 8(1):61–80.
Pallinder-Law, A. (1973). Government in Abe.okuta, 1830-1914, with Special Reference to
the E.gba United Government, 1898-1914. sn].
Poirier, D. J. (1980). Partial observability in bivariate probit models. Journal of Economet-
rics, 12(2):209–217.
40 JAMES FENSKE
Price, G. N. (2003). Economic growth in a cross-section of nonindustrial countries: Does
colonial heritage matter for Africa? Review of Development Economics, 7:478–495.
Roberts, R. and Miers, S. (1988). The end of slavery in Africa, pages 3–68. The end of
slavery in Africa. University of Wisconsin Press.
Rodney, W. (1966). African slavery and other forms of social oppression on the Upper Guinea
coast in the context of the Atlantic slave-trade. Journal of African History, 7(3):431–443.
Ryder, A. F. C. (1965). A reconsideration of the Ife-Benin relationship. Journal of African
History, 6(1):25–37.
Sappington, J. M., Longshore, K. M., and Thompson, D. B. (2007). Quantifying landscape
ruggedness for animal habitat analysis: A case study using bighorn sheep in the mojave
desert. Journal of Wildlife Management, 71(5):1419–1426.
Spear, T. T. (1997). Mountain farmers: Moral economies of land and agricultural develop-
ment in Arusha and Meru. James Currey; University of California Press, Oxford; Berkeley,
Calif.
Tambiah, S. J. and Goody, J. (1973). Bridewealth and Dowry. Cambridge University Press,
Cambridge.
Tertilt, M. (2005). Polygyny, fertility, and savings. Journal of Political Economy,
113(6):1341–1371.
The Author (2010). Land abundance and economic institutions: Egba land and slavery,
1830-1914. Working Paper.
Thornton, J. (1992). Africa and Africans in the Making of the Atlantic World, 1400-1680.
Cambridge University Press.
Uchendu, V. C. (1979). Slaves and slavery in Iboland, Nigeria, pages 121–132. Slavery in
Africa: historical and anthropological perspectives. University of Wisconsin Press.
Wagner, G. (1949). The Bantu of western Kenya. With special reference to the Vugusu and
Logoli. Oxford University Press, London.
Watson, J. L. (1980). Introduction: Slavery as an institution, open and closed systems, pages
1–15. Asian and African systems of slavery. University of California Press, Berkeley and
Los Angeles.
Webb, J. L. A. (1995). Desert Frontier: Ecological and Economic Change Along the Western
Sahel, 1600-1850. University of Wisconsin Press.
Wint, W. and Rogers, D. (2000). Predicted distributions of tsetse in Africa. Report and data-
base prepared by Environmental Research Group Oxford Ltd and TALA Research Group.
LAND ABUNDANCE 41
Appendix A. Data Appendix: Not for Publication
This appendix gives sources and definitions for the geographic variables used and lists
the matches used to connect the ethnic groups from Murdock (1967) to those in Murdock’s
(1959) map.
Table 10: Data Sources
Variable Description Citation Link
Agric. Suit. The GAEZ reports an index be-
tween 0 and 10 combining cli-
mate, soil and terrain slope con-
straints. The “Agric. Suit.”
Measure is this the maximum of
the observed values, minus this
index, divided by the observed
range.
Fischer et al.
(2002)
http://www.
iiasa.ac.at/
Research/LUC/
SAEZ/index.html
Pop. Dens. 1960 Population density in 1960 per
Sq. Km.
UNESCO
(1987)
through
UNEP/GRID-
Sioux Falls
http://na.unep.
net/datasets/
datalist.php
Elevation Elevation in Km. N/A http://epp.
eurostat.ec.
europa.eu/
Precipitation Average annual precipitation
(m). Missing values (due to
differences in resolution between
the data and map) are imputed
using the nearest raster point.
Fischer et al.
(2002)
http://www.
iiasa.ac.at/
Research/LUC/
SAEZ/index.html
Continued on next page
42 JAMES FENSKE
Variable Description Citation Link
Temperature The accumulated temperature
on days with mean daily tem-
perature above 0 C. 55537 is
treated as missing and these
points are dropped before the
join. Missing values imputed us-
ing the nearest raster point.
Fischer et al.
(2002)
http://www.
iiasa.ac.at/
Research/LUC/
SAEZ/index.html
Abs. Latitude Absolute value of the latitude of
the ethnic group’s centroid, re-
ported by ArcMap.
N/A N/A
Dist. to Lake Vic-
toria
Distance, in 1000 Km, from the
ethnic group’s centroid to the
center of Lake Victoria, calcu-
lated using the globdist function
for Stata written by Kenneth Si-
mons.
N/A N/A
Dist. to Coast Average distance from all points
in the ethnic group territory to
the nearest point on the coast,
in decimal degrees, calculated in
ArcMap.
N/A N/A
Malaria Suit. Climatic suitability for malaria
transmission.
Adjuik et al.
(1998)
http://www.
mara.org.za/
lite/download.
htm
Continued on next page
LAND ABUNDANCE 43
Variable Description Citation Link
Tsetse Suit. The raw data is the predicted
presence of tsetse using satellite
imagery on eco-climatic data,
human population, and pre-
dicted cattle and cultivation lev-
els. Because human population
may be endogenous, this is con-
verted into a binary variable (1
if it is greater than 0.5) and re-
gressed as a probit on quadrat-
ics in precipitation, elevation,
temperature, latitude and longi-
tude. The predicted probability
from this probit is used as the
measure of tsetse suitability.
Wint and
Rogers (2000)
http://ergodd.
zoo.ox.ac.uk/
paatdown/index.
htm
Ruggedness This is calculated using the user-
written Vector Ruggedness Mea-
sure script for ArcMap. It “mea-
sures terrain ruggedness as the
variation in three-dimensional
orientation of grid cells within
a neighborhood.” The input
data is the elevation data listed
above, and the neighborhood
size selected is 3, the smallest
possible. Missing values im-
puted using the nearest raster
point.
Sappington
et al. (2007)
http://
arcscripts.
esri.com/
details.asp?
dbid=15423
Continued on next page
44 JAMES FENSKE
Variable Description Citation Link
Population
Weight
This is an estimate of the popu-
lation of the ethnic group, calcu-
lated by summing over the pop-
ulations of cells contained in the
population density data. If more
than one group is assigned to a
single territory, the population
of each group is taken as the sum
of the population within that
territory divided by the number
of groups.
N/A N/A
LAND ABUNDANCE 45
Table 11: Matches
Name in Atlas Name in Map ISO 639-3 Name in Atlas Name in Map ISO 639-3
Alternative Spelling
AULLIMIND AULLIMINDEN ttq KARAMOJON KARAMOJONG kdj
BAFIA FIA ksf KIPSIGIS KIPSIGI kln
BALI LI mhk KURAMA KURAMA, GURE (SE) krh
BAMUM MUM bax LAKA LAKA(ADAMAWA) lak
BANEN NEN baz LENGE HLENGWE cce
BANYANG ANYANG ken MBANDJA BANZA mmz
BASAKOMO BASA bzw NGULU NGURU ngp
BENIAMER AMER amf NGUMBI NGUMBE khu
BIRIFOR BIRIFON bfo NYANKOLE NKOLE nyn
BISA BUSANSI bib PLTONGA TONGA toi
BOMBESA MBESA zms PLAINSBIR BIRA brf
CHAGGA CHAGA jmc/old PLAINSSUK SUK pko
CHAWAI JERAWA, CHAWAI(SW) cch SAPEI SABEI kpz
DAKAKARI BAKAKARI dri SARA SALA sba
FUNGOM FUNGON bfm SHAWIYA SHAWIA shy
FUTAJALON FOUTADJALON fuf SIWANS SIWA siz
GIRIAMA GYRIAMA nyf XHOSA XOSA xho
GURE KURAMA, GURE (SE) krh ZENAGA ZENEGA zen
HILLSUK SUK pko ZINZA SINZA zin
HONA KONA hwo ZUANDE ZUANDE, BATU(E) N/A
Alternative Name
ABRON BRONG abr KAKWA BARI keo
AWUNA GRUNSHI ewe LAKETONGA NYASA tog
BOROROFUL SOKOTO fuv MAMBWE LUNGU mgr
FALASHA KEMANT ahg MBUTI LESE les
GALAB RESHIAT dsh NGONDE NYAKYUSA nyy
HATSA KINDIGA hts RIFFIANS RIF rif
JIMMA JANJERO jnj TURA GURO goa/neb
KAGURU SAGARA kki
Subgroup
SHONA KARANGA sna SOMALI MIJERTEIN som
SIDAMO KAMBATA sid TSWANA NGWATO tsn
Alternate Subgroup
ALAGYA AVIKAM ald SHANGAMA BAKO aiz
KAGORO KATAB kcg UBAMER BAKO aiz
NANKANSE GURENSI gur
Supergroup
Continued on next page
46 JAMES FENSKE
Name in Atlas Name in Map ISO 639-3 Name in Atlas Name in Map ISO 639-3
AFIKPO IBO ibo HAMMAR OMETO amf
ANFILLO MAO myo KANAWA HAUSA hau
ARBORE KONSO arv KASENA GRUNSHI xsm
BANNA OMETO amf MALE OMETO mdy
BASKETO OMETO bst NGONI SENGA ngo
BASSARI TENDA bsc TALLENSI GURENSI gur
BOMVANA XOSA xho TSAMAI KONSO tsb
CONIAGUI TENDA cou VUGUSU LUO luo
DIME OMETO dim YATENGA MOSSI mos
DORSE OMETO doz ZAZZAGAWA HAUSA hau
EFIK IBIBIO efi
Location
ANAGUTA AFUSARE nar LOWIILI BIRIFON N/A
BADITU OMETO N/A MESAKIN KOALIB jle
BODI TOPOTHA mym MORO TALODI mor
BURJI BORAN bji NYARO KOALIB fuj
DJAFUN NAMSHI fub OTORO TAGALI otr
ISALA WABA sil SHAKO KAFA she
KARA KEREWE reg TIRA TALODI tic
KORONGO TUMTUM kgo TIRIKI NANDI ida
KUSASI GURENSI kus TULLISHI NYIMA tey
LALIA KELA lal WODAABE KANURI fuq
Notes:For AWUNA, see Grindal (1972). KANAWA refers to the city of Kano. For VUGUSU, see Wagner
(1949).YATENGA refers to the Mossi capital. ZAZZAGAWA refers to the city of Zaria. Djafun-Bororo is a Fulbe
group.